Pubblicazioni scientifiche
Evolutionary Constraints Guide AlphaFold2 in Predicting Alternative Conformations and Inform Rational Mutation Design
Abstract Investigating structural variability is essential for understanding protein biological functions. Although AlphaFold2 accurately predicts static structures, it fails to capture the full spectrum of functional states. Recent methods have used AlphaFold2 to generate diverse structural ensembles, but they offer limited interpretability and overlook the evolutionary signals underlying the predictions. In this work, we enhance the generation of conformational ensembles and identify sequence patterns that influence the alternative fold predictions for several protein families. Building on prior research that clustered multiple sequence alignments to predict fold-switching states, we introduce a refined clustering strategy that integrates protein language model representations with hierarchical clustering, overcoming limitations of density-based methods. Our strategy effectively identifies high-confidence alternative conformations and generates abundant sequence ensembles, providing a robust framework for applying direct coupling analysis (DCA). Through DCA, we uncover key coevolutionary signals within the clustered alignments, leveraging them to design mutations that stabilize specific conformations, which we validate using alchemical free energy calculations from molecular dynamics. Notably, our method extends beyond fold-switching, effectively capturing a variety of conformational changes. Authors Valerio Piomponi, Alberto Cazzaniga, Francesca Cuturello Journal Journal of Chemical Information and Modelling Publication Date 03/09/2025 Consult the publication
Rational optimization of D3R/GSK-3β dual target-directed ligands as potential treatment for bipolar disorder: Design, synthesis, X-ray crystallography, molecular dynamics simulations, in vitro ADME, and in vivo pharmacokinetic studies
Abstract Bipolar disorder is a complex neuropsychiatric condition with a significant unmet medical need, as current treatments lack disease-modifying properties and multimodal therapeutic effects. To overcome the limitations of single-target drugs, we designed dual-target ligands that combine partial agonism at the dopamine D3 receptor (D3R) with inhibition of glycogen synthase kinase-3β (GSK-3β). We previously identified ARN24161 (1) as a promising prototype, demonstrating partial agonism at D3R (EC50 = 10.1 nM, % Eff. = 26.3) and GSK-3β inhibition (IC50 = 561 nM). However, its drug-like properties remained suboptimal. To optimize this compound, we initiated a multidisciplinary refinement campaign, leveraging computational modeling and crystallographic data to fine-tune the balance between D3R and GSK-3β activity, reduce P-glycoprotein (P-gp) affinity, and improve the pharmacokinetic profile. This effort led to the identification of ARN25297 (5), a moderately balanced dual-target ligand that exhibits partial agonism at D3R (EC50 = 13.1 nM, % Eff. = 17.1) and potent GSK-3β inhibition (IC50 = 47.0 nM). Notably, ARN25657 (16) emerged as the most well-balanced candidate, demonstrating enhanced D3R partial agonism (EC50 = 15.2 nM, % Eff. = 37.7) alongside strong GSK-3β inhibition (IC50 = 19.3 nM). Compound 16 also exhibited the lowest P-gp inhibition and significant improvements in in vitro ADME properties compared to prototype 1, while maintaining a balanced dual target profile. Although the PK profile of 16 remained comparable to that of prototype 1, these findings lay the groundwork for further lead optimization and structural refinement, driving future in vivo proof-of-concept toward innovative therapeutic strategies for bipolar disorder and related neuropsychiatric conditions. Autori RMC Di Martino, D Russo, I Penna, Andrea Dalle Vedove, R Spabnuolo, G Ottonello, M Summa, J Desantis, A Valeri, L Pruccoli, SK Tripathi, A Tarozzi, Paola Storici, S Girotto, R Bertorelli, A Armirotti, G Cruciani, T Bandiera, A Cavalli, G Bottegoni Rivista European Journal of Medicinal Chemistry Data di pubblicazione 25/06/2025 Consulta la pubblicazione
FL30: an epidermal growth factor kinase inhibitor overcoming T790M and C797S mutations through unique conformational modulation mechanism
Abstract Tyrosine kinase inhibitors (TKIs) targeting the oncogene Epidermal Growth Factor Receptor (EGFR) are widely used in the treatment of non-small cell lung cancer (NSCLC). In this context, the introduction of fourth-generation TKIs has significantly advanced targeted therapy for T790M and C797S EGFR mutations. Current therapeutic strategies are increasingly focusing on the design of orthoallosteric TKIs, which have shown promise in stabilizing the inactive conformation of mutated EGFR. In this context, we report the discovery of FL30, a small molecule with a flavone core that exhibits nanomolar potency against the EGFR-L858R/T790M mutation, even in the presence of the C797S mutation. The IC50 comparable to the Osimertinib – one of the most renowned EGFR-TKIs – emphasizes the remarkable success of the design approach. In NSCLC models, FL30 effectively inhibits cancer growth and EGFR phosphorylation selectively in cells with the EGFR mutations. Kinetic studies, molecular modeling, and Plasmon Internal Reflection Surface-Enhanced Infrared Absorption (PIR-SEIRA) microscopy suggests that FL30 binds to the orthosteric site while inducing the transition of the mutant EGFR toward an inactive-like state. These findings highlight FL30’s potential for further optimization and propose a novel approach for developing targeted therapies that combine orthosteric binding with allosteric modulation. Autori Elena Romagnoli, Emiliano Laudadio, Giovanna Mobbili, Leonardo Sorci, Giovanni Birarda, Federica Piccirilli, Lisa Vaccari, Hendrik Vondracek, Brenad Romaldi, Massimo Marcaccio, Paola Storici, Marta Semrau, Roberta Galeazzi, Andrea Toma, Vincenzo Aglieri, Pierluigi Stipa, Tatiana Armeni, Cristina Minnelli Rivista International Journal of Biological Macromolecules Data di pubblicazione 21/06/2025 Consulta la pubblicazione
Unsupervised Domain Classification of AlphaFold2-Predicted Protein Structures
Abstract The release of the AlphaFold database, which contains 214 million predicted protein structures, represents a major leap forward for proteomics and its applications. However, the lack of comprehensive protein annotation limits its accessibility and usability. Here, we present DPCstruct, an unsupervised clustering algorithm designed to provide domain-level classification of protein structures. Using structural predictions from AlphaFold2 and comprehensive all-against-all local alignments from Foldseek, DPCstruct identifies and groups recurrent structural motifs into domain clusters. When applied to the Foldseek Cluster database, a representative set of proteins from the AlphaFoldDB, DPCstruct successfully recovers the majority of protein folds catalogued in established databases such as SCOP and CATH. Out of the 28 246 clusters identified by DPCstruct, 24% have no structural or sequence similarity to known protein families. Supported by a modular and efficient implementation, classifying 15 million entries in less than 48 h, DPCstruct is well suited for large-scale proteomics and metagenomics applications. It also facilitates the rapid incorporation of updates from the latest structural prediction tools, ensuring that the classification remains up-to-date. The DPCstruct pipeline and associated database are freely available in a dedicated repository, enhancing the navigation of the AlphaFoldDB through domain annotations and enabling rapid classification of other protein datasets. Authors Federico Barone, Alessandro Laio, Marco Punta, Stefano Cozzini, Alessio Ansuini, Alberto Cazzaniga Journal Physical Review X Life Publication Date 27/05/2025 Consult the publication
Persistent Topological Features in Large Language Models
Abstract Understanding the decision-making processes of large language models is critical given their widespread applications. To achieve this, we aim to connect a formal mathematical framework—zigzag persistence from topological data analysis —with practical and easily applicable algorithms. Zigzag persistence is particularly effective for characterizing data as it dynamically transforms across model layers. Within this framework, we introduce topological descriptors that measure how topological features, -dimensional holes, persist and evolve throughout the layers. Unlike methods that assess each layer individually and then aggregate the results, our approach directly tracks the full evolutionary path of these features. This offers a statistical perspective on how prompts are rearranged and their relative positions changed in the representation space, providing insights into the system’s operation as an integrated whole. To demonstrate the expressivity and applicability of our framework, we highlight how sensitive these descriptors are to different models and a variety of datasets. As a showcase application to a downstream task, we use zigzag persistence to establish a criterion for layer pruning, achieving results comparable to state-of-the-art methods while preserving the system-level perspective. Authors Yuri Gardinazzi, Karthik Viswanathan, Giada Panerai, Alessio Ansuini, Alberto Cazzaniga, Matteo Biagetti Journal International Conference of Machine Learning (ICML) 2025 Publication Date 01/05/2025 Consult the publication
Heterologous prime–boost Zika virus vaccination induces comprehensive humoral and cellular immunity in mouse models
Abstract Zika virus (ZIKV) remained poorly studied until an outbreak in 2015 linked the virus to severe neurological disorders and congenital malformations. Currently, there are no antiviral drugs or vaccines available. We have previously demonstrated that a simian adenovirus vector vaccine (ChAdOx1 prMEΔTM) and a virus-like particle-based vaccine bearing E proteins locked in covalent dimers (VLP-cvD) are effective against ZIKV infection in animal challenge models. In this study, we further explored the efficacy of these vaccines, either individually or in combination, using a heterologous prime and boost vaccination strategy in mouse challenge models. Although the individual vaccines provided good protection levels, the heterologous prime–boost vaccination regimen (ChAdOx1 prMEΔTM followed by VLP-cvD) offered the most effective protection. This regimen elicited a strong cellular response and high levels of neutralising antibodies, which were attributed to ChAdOx1 prMEΔTM and VLP-cvD, respectively. Our findings support the use of combined vaccine technologies and offer valuable insights into the multifactorial protection achievable through heterologous vaccination. These results have important implications for the development of effective vaccination strategies against ZIKV and other emerging viruses. Autori Giuditta De Lorenzo, Rapeepat Tandavanitj, Lorena Preciado-Llanes, Ricardo Sanchez-Velazquez, Raissa Prado Rocha, Young Chan Kim, Arturo Reyes-Sandoval, Arvind H Patel. Rivista Frontiers in Immunology Data di pubblicazione 25/04/2025 Consulta la pubblicazione
Broadband terahertz signatures and vibrations of Phe–Phe peptide and its fibrils
Abstract In recent years, peptide-based nanomaterials have gained significant attention in drug discovery due to their biocompatibility and promising functionality in biophysical processes. This current study employs terahertz (THz) spectroscopy and density functional theory (DFT) to investigate the vibrational properties of the phenylalanine dipeptide (Phe–Phe), a building block with notable self-assembling properties and potential applications in drug delivery and nanostructured biomaterials. The dynamics of proteins and biomolecules occurring on the picosecond timescale can be probed by THz spectroscopy and is closely related to their functionality. Here, we investigate the low-frequency vibrational modes of Phe–Phe under two different conditions, as crystalline commercial powder and post-self-assembled powder, in the 0.2–4 THz range. The refractive index of pure Phe–Phe, as evaluated from THz spectroscopy, is approximately 1.47, and the THz absorption peaks are observed at 0.55, 0.83, 1.13, 1.40, 1.68, 2.18, 2.71, 3.00, and 3.33 THz. Potential energy distribution (PED) analysis provides a detailed assignment of the observed modes and identifies characteristic vibrational features. The novel thin bi-layer approach for sample preparation employed here proved to be effective in terms of signal-to-noise ratio and in eliminating artifacts possibly originating from the host material. This combined experimental–computational approach not only offers valuable insights into the conformational flexibility and self-assembly potential of Phe–Phe, but also underscores the efficacy of THz spectroscopy and DFT analysis for studying the vibrational properties of peptides, with implications for biophysics, nanotechnology, and biochemistry. Moreover, this study highlights the impact of intermolecular interactions on the vibrational spectra by comparing crystalline powder and self-assembled peptide powder. Autori Rajat Kumar, Federica Piccirilli, Paola Di Pietro, Johannes Schmidt, Giovanni Birarda, Lisa Vaccari, Andrea Perucchi, Prasanta Kumar Datta Rivista Analyst Data di pubblicazione 24/04/2025 Consulta la pubblicazione
ResiDual Transformer Alignment with Spectral Decomposition
Abstract When examined through the lens of their residual streams, a puzzling property emerges in transformer networks: residual contributions (e.g., attention heads) sometimes specialize in specific tasks or input attributes. In this paper, we analyze this phenomenon in vision transformers, focusing on the spectral geometry of residuals, and explore its implications for modality alignment in vision-language models. First, we link it to the intrinsically low-dimensional structure of visual head representations, zooming into their principal components and showing that they encode specialized roles across a wide variety of input data distributions. Then, we analyze the effect of head specialization in multimodal models, focusing on how improved alignment between text and specialized heads impacts zero-shot classification performance. This specialization-performance link consistently holds across diverse pre-training data, network sizes, and objectives, demonstrating a powerful new mechanism for boosting zero-shot classification through targeted alignment. Ultimately, we translate these insights into actionable terms by introducing ResiDual, a technique for spectral alignment of the residual stream. Much like panning for gold, it lets the noise from irrelevant unit principal components (i.e., attributes) wash away to amplify task-relevant ones. Remarkably, this dual perspective on modality alignment yields fine-tuning level performance on different data distributions while modelling an extremely interpretable and parameter-efficient transformation, as we extensively show on 70 pre-trained network-dataset combinations (7 models, 10 datasets). Authors Lorenzo Basile, Valentino Maiorca, Luca Bartolussi, Emanuele Rodolà, Francesco Locatello Journal Transmissions in Machine Learning Research Publication Date 12/04/2025 Consult the publication
Activation Patching for Interpretable Steering in Music Generation
Abstract Understanding how large audio models represent music, and using that understanding to steer generation, is both challenging and underexplored. Inspired by mechanistic interpretability in language models, where direction vectors in transformer residual streams are key to model analysis and control, we investigate similar techniques in the audio domain. This paper presents the first study of latent direction vectors in large audio models and their use for continuous control of musical attributes in text-to-music generation. Focusing on binary concepts like tempo (fast vs. slow) and timbre (bright vs. dark), we compute steering vectors using the difference-in-means method on curated prompt sets. These vectors, scaled by a coefficient and injected into intermediate activations, allow fine-grained modulation of specific musical traits while preserving overall audio quality. We analyze the effect of steering strength, compare injection strategies, and identify layers with the greatest influence. Our findings highlight the promise of direction-based steering as a more mechanistic and interpretable approach to controllable music generation. Authors Simone Facchiano, Giorgio Strano, Donato Crisostomi, Irene Tallini, Tommaso Mencattini, Fabio Galasso, Emanuele Rodolà Journal Preprint Publication Date 06/04/2025 Consult the publication
Eco-friendly NaCl glycerol-based deep eutectic electrolyte for high-voltage electrochemical double layer capacitor
Abstract Herein, we propose eco-friendly electrolytes based on sodium chloride as a hydrogen bond acceptor and glycerol as a hydrogen bond donor, as alternatives to toxic, flammable and unsustainable electrolytes commonly used in electrochemical energy storage systems. By means of an in-depth multi-technique investigation, including Raman and FT-FIR spectroscopy, of the formulated electrolytes, we point out the effect of the structuring of the system on the transport and electrochemical properties. The 1 : 10 molar ratio mixture proves to be a deep eutectic solvent (DES), showing good room temperature ionic conductivity (0.186 mS cm−1) and electrochemical stability (≈3 V). When implemented as electrolyte in an activated-carbon electrochemical double layer capacitor, this DES exhibits superior performance compared to mixtures with different molar ratio and those containing ethylene glycol as the hydrogen bond donor, with a high operational voltage (2.6 V), a specific capacitance of 14.1 F g−1, and a remarkable cycling stability. These findings highlight the potential of glycerol-based DESs as alternative electrolytes for sustainable electrochemical energy storage applications. Autori Daniele Motta, Alessandro Damin, Hamideh Darjazi, Stefano Nejrotti, Federica Piccirilli, Giovanni Birarda, Claudia Barolo, Claudio Gerbaldi, Giuseppe Antonio Elia, Matteo Bonomo Rivista Green Chemistry Data di pubblicazione 31/03/2025 Consulta la pubblicazione
Cobalt recycling patents dataset selected using ‘green’ classification codes: Focus on the nickel manganese cobalt (NMC) batteries recycling
Abstract Il Critical Raw Materials Act è stato emanato per affrontare i rischi percepiti riguardanti le catene di approvvigionamento di materie prime critiche alla luce della crescita prevista della domanda. Il cobalto è una delle materie prime critiche per l’UE: svolge un ruolo centrale nella transizione verso un’economia a basse emissioni di carbonio ed è cruciale per le batterie che alimentano i veicoli elettrici. L’industria del cobalto è importante non solo per la transizione verde, ma anche perché crea opportunità di lavoro, dato che l’Europa è la terza regione più vasta nella catena del valore globale del cobalto. Il mercato globale dei veicoli elettrici a batteria (BEV) è in continuo aumento, il che comporta una consistente domanda di materiali per la produzione di batterie agli ioni di litio (LIB). In particolare, le batterie al litio nichel manganese cobalto (NMC) sono uno dei tipi di batterie principali utilizzati nei BEV e il riciclo dei materiali dalle batterie usate per produrre nuove batterie può mitigare il rischio di approvvigionamento di materiali. Il presente studio concerne la creazione e il confronto di due dataset di brevetti riguardanti il riciclo del cobalto, rispettivamente ottenuti senza (dataset 1) e con (dataset 2) l’aggiunta di un codice di classificazione CPC specifico ‘verde’ ad una vasta lista di codici di classificazione riguardanti il riciclo dei materiali. In particolare, il CPC Y02P10/20 determina un miglioramento significativo dell’esaustività della ricerca riguardante il riciclaggio del cobalto nel dataset 2, d’altra parte potrebbe anche generare problemi evidenziati da una successiva selezione di quei risultati che in linea di principio sono focalizzati sul recupero delle batterie NMC ma caratterizzati da diversi falsi positivi. Pertanto, si raccomanda una strategia di affinamento per rendere l’accuratezza del dataset NMC comparabile a quella che caratterizza i risultati NMC del dataset 1 ottenuti senza applicare alcun criterio di affinamento. Author Riccardo Priore Journal Data in Brief, March 2024, 105519 Date Available online 13/03/2024 Consult the paper
A supported lipid bilayer to model solid-ordered membrane domains
Abstract Membrane models are widely used to mimic the behaviour of native plasma membranes and to simulate interactions occurring at their interface. Such models can be built up with different molecular compositions, ranging from single phospholipids to more complex, heterogeneous mixtures of phospho- and sphingo-lipids, possibly enriched with cholesterol and proteins. In particular, mixing different lipids and cholesterol is instrumental to promote the formation of phase-separated, ordered domains, which resemble the structure of lipid rafts, specialized functional domains of real membranes. According to the specific lipid composition, physical characteristics of the rafts can be tuned, such as fluidity, strongly related to membrane biological activity. Here, we introduce a novel three-component membrane model constituted by the mixing of a saturated phospholipid, 1,2-dimyristoyl-sn–glycero-3-phosphocholine (DMPC), sphingomyelin and cholesterol to mimic the presence of solid ordered rafts and to study their behaviour. Differential scanning calorimetry, neutron reflectometry, and atomic force microscopy were synergistically applied to gain information on the membrane’s transverse and lateral organization, as well as on its thermotropic behaviour. The membrane model benefits from the use of DMPC, a lipid (i) characterized by an accessible transition temperature; (ii) saturated; (iii) fluid at physiological temperature and (iv) commercially available in both protiated and deuterated forms. The proposed model, along with the wide range of biophysical techniques employed, constitutes an ideal system to study the molecular mechanisms and the physical properties that govern membrane functions, such as molecular signalling and membrane trafficking. Autori Sally Helmy, Paola Brocca, Alexandros Koutsioubas, Stephen C.L. Hall, Luca Puricelli, Pietro Parisse, Loredana Casalis, Valeria Rondelli Rivista Journal of Colloid and Interface Science Data di pubblicazione 14/03/2025 Consulta la pubblicazione